Biology subjects

Stolovitzky, G. A.

Publications and source records attributed to Stolovitzky, G. A..

6 recordsLinked to original sources

bioRxiv2021

Correlated Drug Action as a Baseline Additivity Model for Combination Cancer Therapy in Patient Cohorts and Cell Cultures

Identifying and characterizing the effect of combination therapies is of paramount importance in various diseases, including cancer. Various competing null models have been proposed to serve as baselines against which to compare the effect of drug combinations. In this work, we introduce Correlated Drug Action (CDA), a baseline model for the study of drug combinations in both cell cultures and in patient populations. CDA assumes that the efficacy of pairs of drugs to be used in a combination may be correlated, that is, if the efficacy of a drug in a given patient or cell is high (low), then the efficacy of the other drug may also be high (low) in the same patient or cell. Our model can be used in the temporal domain (temporal CDA or tCDA) to explain survival curves in patient populations, and in the dose domain (dose CDA or dCDA), to explain dose-response curves in cell cultures. At the level of clinical trials, we demonstrate tCDAs utility in identifying possibly synergistic combinations and cases where the combination can be explained in terms of the monotherapies. At the level of cells in culture, dCDA generalizes null models such as Bliss independence, the Highest Single Agent model, the dose equivalence principle, and is consistent with what should be expected in sham combinations. We demonstrate the applicability of dCDA in assessing combinations in experimental MCF7 cell-line data by introducing a new metric, the Excess over CDA (EOCDA).

bioinformatics↗

bioRxiv2020

DreamAI: algorithm for the imputation of proteomics data

Deep proteomics profiling using labeled LC-MS/MS experiments has been proven to be powerful to study complex diseases. However, due to the dynamic nature of the discovery mass spectrometry, the generated data contain a substantial fraction of missing values. This poses great challenges for data analyses, as many tools, especially those for high dimensional data, cannot deal with missing values directly. To address this problem, the NCI-CPTAC Proteogenomics DREAM Challenge was carried out to develop effective imputation algorithms for labeled LC-MS/MS proteomics data through crowd learning. The final resulting algorithm, DreamAI, is based on an ensemble of six different imputation methods. The imputation accuracy of DreamAI, as measured by Pearson correlation, is about 15%-50% greater than existing tools among less abundant proteins, which are more vulnerable to be missed in proteomics data sets. This new tool notably enhances data analysis capabilities in proteomics research.

bioinformatics↗

bioRxiv2020

Crowdsourcing assessment of maternal blood multi-omics for predicting gestational age and preterm birth

Identification of pregnancies at risk of preterm birth (PTB), the leading cause of newborn deaths, remains challenging given the syndromic nature of the disease. We report a longitudinal multi-omics study coupled with a DREAM challenge to develop predictive models of PTB. We found that whole blood gene expression predicts ultrasound-based gestational ages in normal and complicated pregnancies (r=0.83), as well as the delivery date in normal pregnancies (r=0.86), with an accuracy comparable to ultrasound. However, unlike the latter, transcriptomic data collected at <37 weeks of gestation predicted the delivery date of one third of spontaneous (sPTB) cases within 2 weeks of the actual date. Based on samples collected before 33 weeks in asymptomatic women we found expression changes preceding preterm prelabor rupture of the membranes that were consistent across time points and cohorts, involving, among others, leukocyte-mediated immunity. Plasma proteomic random forests predicted sPTB with higher accuracy and earlier in pregnancy than whole blood transcriptomic models (e.g. AUROC=0.76 vs. AUROC=0.6 at 27-33 weeks of gestation).

bioinformatics↗

bioRxiv2020

Unannotated small RNA clusters in circulating extracellular vesicles detect early stage liver cancer

BackgroundHepatocellular carcinoma (HCC) is among the deadliest malignancies and surveillance tools for early detection are suboptimal. Extracellular vesicles (EVs) have gained increasing scientific interest due to their involvement in tumor initiation and metastasis, however, most extracellular RNA (exRNA) biomarker studies are limited to annotated genomic regions. MethodsEVs were isolated with ultracentrifugation and nanoDLD and quality assessed by electron microscopy, immunoblotting, nanoparticle tracking, and deconvolution analysis. We performed genome-wide small exRNA sequencing, including unannotated transcripts. We identified small RNA clusters (smRCs) and delineated their key genomic features across biospecimens (blood, urine, tissue) and EV isolation techniques. A 3-smRC signature for early HCC detection was trained and validated in two independent cohorts. ResultsEV-derived smRCs were dominated by uncharacterized, unannotated small RNA and uniformly tiled across the genome with a consensus sequence of 20bp. A 3-smRC signature was significantly overexpressed in circulating EVs of HCC patients compared to controls at risk or patients with non-HCC malignancies (p<0.01, n=157). An independent validation in a phase 2 biomarker study revealed 86% sensitivity and 91% specificity for the detection of early HCC from controls at risk (i.e. cirrhosis or chronic liver disease, n=209) (positive predictive value (PPV): 89%, area under the ROC curve [AUC]: 0.87). The 3-smRC signature was independent of alpha-fetoprotein (p<0.0001) and a composite model yielded an increased AUC of 0.93 (sensitivity: 85%, specificity: 94%, PPV: 95%). ConclusionAn exRNA-based 3-smRC signature from plasma detects early stage HCC, which directly leads to the prospect of a minimally-invasive, blood-only, operator-independent surveillance biomarker. One sentence summaryWe employ a novel, data-driven approach to identify and characterize small RNA clusters from unannotated loci in extracellular vesicle-derived RNA across different cancer types, isolation techniques, and biofluids, facilitating discovery of a robust biomarker for detection of early stage liver cancer.

cancer biology↗

bioRxiv2020

Crowdsourced mapping of unexplored target space of kinase inhibitors

Despite decades of intensive search for compounds that modulate the activity of particular targets, there are currently small-molecules available only for a small proportion of the human proteome. Effective approaches are therefore required to map the massive space of unexplored compound-target interactions for novel and potent activities. Here, we carried out a crowdsourced benchmarking of predictive models for kinase inhibitor potencies across multiple kinase families using unpublished bioactivity data. The top-performing predictions were based on kernel learning, gradient boosting and deep learning, and their ensemble resulted in predictive accuracy exceeding that of kinase activity assays. We then made new experiments based on the model predictions, which further improved the accuracy of experimental mapping efforts and identified unexpected potencies even for under-studied kinases. The open-source algorithms together with the novel bioactivities between 95 compounds and 295 kinases provide a resource for benchmarking new prediction algorithms and for extending the druggable kinome.

biochemistry↗

bioRxiv2017

Prediction of Molecular Mechanisms of Breast Cancer Metastasis

Metastasis -the spread of cancer to other parts of the body- causes 90% of cancer deaths, underlies major health complications in cancer patients and renders most cancers incurable. Unfortunately, the molecular mechanisms underlying the process are poorly understood and therapeutics to block it remain elusive. Here, we present a computational technique for scanning genome-scale regulatory networks for potential genes associated with metastasis. First, we demonstrate that in the breast cancer cell line MCF7, the commonly dysregulated cancer biomarkers TP53, ERBB2, ESR1 and PGR are closely connected to known metastasis genes with a significant proportion being 2nd degree neighbors of a given biomarker. Next, we identify genes whose 2nd degree neighbors are connected in a similar manner to these biomarkers. Consequently, these are referred to as metastasis associated genes or MAGs. We identify 190 genes that are TP53-MAGs, 22 ERBB2-MAGs, 240 ESR1-MAGs and 84 PGR-MAGs (FDR adjusted P <0.001). Analysis of the MAGs reveals statistically significant enrichment with biological functions previously associated with metastasis including the extracellular matrix (ECM) receptor interaction, focal adhesion, cytokine-cytokine receptor interaction and chemokine signaling. The biological significance of MAGs is further supported by their enrichment with experimentally validated binding sites for transcription factors that regulate metastasis, for example BACH1- a master regulator of breast cancer metastasis to bone. The predicted MAGs are also clinically relevant as therapeutic targets for metastasis blocking agents. Specifically, genes that are perturbed by drugs and miRNAs that influence metastasis are enriched with MAGs. Furthermore, some MAGs are associated with patient survival and provide insights into the proclivity for breast cancer subtypes to preferentially spread to specific organs. The results of this study imply that aberrations in primary tumors may constrict metastasis trajectories. This could enable the prediction of organ specific metastases based on aberrations in the primary tumor and lay a foundation for future studies on individualized or personalized models of metastasis. The approach is potentially scalable across other cancers and has clinical implications.

cancer biology↗